Nowadays, the application of data caching in mobile edge computing networks is exponentially increasing as a high-speed data storage layer using deep learning (DL) approaches. This paper presents an DL-based resource allocation approach to find the optimal topology for cache-enabled backhaul networks. In the practical scenarios, a numerous of radial configurations of test systems have been applied for training stages. This paper also applied the continuation power flow analysis to achieve the maximum load limit in which the power of macro base stations with the caches of different sizes is provided through either smart grid network or renewable power systems. To increase the power efficiency index of this approach the power sharing capability was enabled among different layer of network components through smart grids. In order to obtain the optimal solution, the DL-based mathematical problem is reformulated into a neural weighting model considering convergence conditions and Lyapunov stability of the mobile-edge-computing under Karush–Kuhn–Tucker optimality constraints. The mathematical analysis and simulation results demonstrate that the performance of the proposed algorithm is better than other energy efficiency algorithms. The proposed approach can effectively increase the total system throughput and network’s utility in addition to guarantee user fairness index.

如今，作为使用深度学习 (DL) 方法的高速数据存储层，数据缓存在移动边缘计算网络中的应用呈指数增长。本文提出了一种基于 DL 的资源分配方法，以找到启用缓存的回程网络的最佳拓扑。在实际场景中，许多测试系统的径向配置已应用于训练阶段。本文还应用连续潮流分析来实现最大负载限制，其中具有不同大小缓存的宏基站的功率通过智能电网网络或可再生电力系统提供。为了提高这种方法的电源效率指标，通过智能电网在不同层的网络组件之间启用了电源共享功能。为了获得最优解，将基于 DL 的数学问题重新表述为考虑收敛条件和 Karush-Kuhn-Tucker 最优约束下移动边缘计算的 Lyapunov 稳定性的神经加权模型。数学分析和仿真结果表明，该算法的性能优于其他能效算法。所提出的方法除了保证用户公平性指标外，还可以有效地提高系统总吞吐量和网络的效用。数学分析和仿真结果表明，该算法的性能优于其他能效算法。所提出的方法除了保证用户公平性指标外，还可以有效地提高系统总吞吐量和网络的效用。数学分析和仿真结果表明，该算法的性能优于其他能效算法。所提出的方法除了保证用户公平性指标外，还可以有效地提高系统总吞吐量和网络的效用。